Hangzhou’s open-model star steps onto the listing track

On September 9, 2026, Reuters reported that DeepSeek has tapped CITIC Securities to prepare for an IPO on Shanghai’s tech-focused STAR Market. China’s 21st Century Business Herald later confirmed with sources that CITIC has entered due diligence — though a formal listing-tutoring agreement is not signed yet. Timing, raise size, and a final valuation target remain open, but the signal is loud: one of the world’s most-watched open-weight labs is lining up for a domestic public listing.

What we actually know

Strip away the rumor fog and the confirmed-or-near-confirmed pieces look like this:

  • Banker: CITIC Securities is in the room for STAR Market prep (Reuters exclusive; local confirmation of due diligence).
  • Venue: Shanghai STAR Market (Sci-Tech Innovation Board), not a U.S. ADR path.
  • Cadence: People familiar with the talks say DeepSeek aims to begin the IPO process this year. That is a process start, not a priced debut date.
  • Capital context: Reuters has previously reported an ongoing private round that would value DeepSeek around $75 billion. A June settlement raised about $7.4 billion at a post-money valuation north of $50 billion.
  • Why the money: Compute, model work, talent retention, and in-house chips — the same cost stack every frontier lab is racing.

Tencent Tech / 36Kr reporting adds operating color that has not been independently audited in this write-up: roughly ¥475 million revenue in the first seven months of 2026 (about 10× full-year 2025), alongside a ¥715 million net loss in the same window, plus a headcount near 300 with a fresh ~150 engineer hiring wave on agent infrastructure. Treat those operating numbers as reported claims until DeepSeek files.

Pro Tip

For creators, the “IPO” headline matters less than the capital behind open weights. A STAR Market path can unlock longer-horizon compute and hiring — which is what keeps models like DeepSeek’s stack shipping for local fine-tunes, agents, and API remixes.

Why this matters for gen-AI builders

DeepSeek is not a distant chip story. Its open-weight releases reshaped how indie studios, research labs, and tool vendors budget inference and fine-tuning. A domestic listing prep:

  1. Signals durability — public-market discipline and STAR disclosure norms force clearer governance and funding plans.
  2. Funds the stack — GPUs, custom silicon, and agent engineering are expensive; public capital is one way to keep shipping open (or open-adjacent) models.
  3. Shifts the competitive map — U.S. and Chinese labs already race on capability; capitalization races now sit beside them.

None of that guarantees weights stay as open as today’s community expects. Listings create board pressure. Watch product licenses and API terms as carefully as the prospectus.

What to watch next

  • Formal tutoring agreement and STAR application filing (process milestones beat valuation gossip).
  • Whether the private round closes near the reported $75B mark.
  • Chip and agent hiring outcomes — DeepSeek’s September recruiting push is already public.
  • Any change in open-weight release cadence once bankers are fully in the loop.

Sunny takeaway: open-model gravity just got a capital-markets chapter. Keep building on the models you love — and keep a bookmark on DeepSeek’s filing trail. Primary reporting to start with: Reuters exclusive and 21st Century Business Herald confirmation.

—Aurelia ♡